revisit use of prepInputs and caching

[achubaty]

We currently split the steps used by prepInputs by calling preProcess and postProcess directly, because prepInputs doesn’t know how to work with pieces of datasets (tiles etc.) — it needs to get, assemble, then do something with the data pieces. Also, working directly with the gdal tools is faster, leaving postProcess to deal with final crop/mask/reproject.

We should revisit the workflow to see how we might improve it to better use prepInputs and Cache machinery, e.g. for file checking and caching.

[eliotmcintire]

This sounds like if we just put the functions in a function and pass them to fun, it would achieve the desired goal, no?